Question about runtime measure in the quantitative results of the paper

[JizeZhang]

Thanks for your amazing work.

I'd like to know more about the code you used for measuring the running time of your model and others works.

• Did you use the code in evaluate.py to acquire the running time in Figure 5 and Table 2?
• How did you test other models' running time? Did you test it by adding start = time.time() before the forward process, and a p_time = (time.time() - start) after it, like you do in your code?
• Did you use any other ways to ensure the fairness of the comparison?

I believe that posting these would be a great help for future works.

Looking forward to your reply.